Investigating State-of-the-Art Machine Learning Approaches in Vegetation Analysis through Earth Observation Data

Dear Colleagues,

Vegetation analysis and mapping is a critical component of monitoring the earth’s ecosystems and understanding the impact of environmental changes on biodiversity and ecosystem services. Accurate vegetation mapping enables researchers and managers to identify and track changes in vegetation cover over time, detect the onset of ecosystem degradation, and identify areas needing restoration or conservation. Furthermore, the analysis of vegetation data provides critical information for climate change research, land-use planning, and agricultural management. Remote sensing has revolutionized vegetation mapping and trend analysis, providing data with different spatial and spectral resolutions on vegetation cover at global scales. However, the accurate interpretation of remote sensing data requires advanced analytical techniques that can handle the complexity and scale of the data.

In recent years, machine learning (ML) techniques have attracted considerable attention for their effectiveness in producing robust vegetation cover maps. With substantial advancements in the field, machine learning is poised to maintain a pivotal role in analysing sizeable remote sensing datasets, incorporating information from varied data sources to extract valuable spatial patterns, for example, support vector machines, decision tree classifiers, Random Forest classifiers, normal Gaussian Baye, an ensemble method utilising various classification methods, or ensemble methods to produce vegetation classification maps. An ensemble classifier is a machine learning technique that combines the predictions of multiple base classifiers to create a more accurate and robust final prediction. The idea is that by aggregating the outputs of several models, the ensemble can overcome the limitations of individual classifiers and improve the overall performance. The strength of an ensemble classifier comes from the diversity of its base models. These models can be different algorithms (e.g., decision trees, support vector machines) or the same algorithm trained on different subsets of data or with different hyperparameters.

The Special Issue, however, also welcomes the application of Convolutional Neural Networks (CNNs) for vegetation mapping. CNNs are a type of deep learning algorithm that has emerged as a powerful approach for analysing remote sensing data and extracting valuable information about vegetation patterns and dynamics. CNNs enable researchers to extract complex features from large-scale remote sensing datasets, providing critical insights into vegetation distribution, composition, and dynamics. The ability of CNNs to accurately classify vegetation types and detect changes in vegetation cover over time can transform our understanding of global vegetation dynamics and its response to environmental changes. This Issue is not limited to the CNN method, and is accepting manuscripts based on other deep learning (DL) methods like recurrent neural networks (RNNs) and deep reinforcement learning for analysis in vegetation dynamics.

The forthcoming Special Issue (SI) aims to highlight the latest developments and applications of ML or deep learning methods in vegetation remote sensing. The SI is open to all kinds of manuscripts, including original research articles, review articles, etc., with the added value of using time series remote sensing data in all aspects regarding the mapping, change detection, trend analysis, and studies of the drivers of vegetation change in all ecosystems using CNNs. Some suggested themes and topics for submission include, but are not limited to, the following:

• DL/other deep learning architectures for vegetation remote sensing:
• Novel CNN/DL architectures designed explicitly for vegetation classification, segmentation, or change detection.
• Comparative studies of different CNN/DL architectures for vegetation remote sensing.
• Applications of DL/ML in vegetation remote sensing:
• Use of DL/ML for vegetation mapping, classification, and segmentation in different regions and ecosystems.
• Analysis of DL/ML performance compared to traditional remote sensing methods in vegetation mapping, monitoring, and trend analysis.
• Retrieving time series of biophysical parameters for vegetation monitoring using DL/ML.
• Integration of DL/ML with other remote sensing data sources, such as LiDAR or hyperspectral data, to improve vegetation mapping accuracy.
• DL/ML for monitoring vegetation dynamics:
• Development of time-series CNN models for vegetation change monitoring and trend analysis.
• Analysis of the spatio-temporal patterns of vegetation changes using DL / ML.
• DL/ML for addressing critical challenges in vegetation remote sensing:
• Use of DL/ML  for the accurate classification of mixed pixel areas, such as urban or agricultural landscapes.
• Response of vegetation dynamics to climatic variables change.
• Analysis of the effects of different data preprocessing techniques on the performance of DL/ML in vegetation remote sensing.
• Development of DL/ML-based techniques for handling missing data in vegetation remote sensing datasets.
• Application of Ensemble algorithms in the field of vegetation dynamics:
• Theoretical architecture and the implementation of ensemble learning algorithms for vegetation mapping, monitoring, or trend analysis.
• Ensemble method to deal with sparse or limited training samples.
• Prediction or classification.
• Comparisons among ensemble methods with conventional machine learning methods for vegetation mapping, monitoring, or predicting.
• Integration of deep learning techniques into ensemble architecture to improve mapping accuracy.

Dr. Abolfazl Abdollahi
Dr. Chandrama Sarker
Guest Editors

Manuscript Submission Information

Manuscripts should be submitted online at www.mdpi.com by registering and logging in to this website. Once you are registered, click here to go to the submission form. Manuscripts can be submitted until the deadline. All submissions that pass pre-check are peer-reviewed. Accepted papers will be published continuously in the journal (as soon as accepted) and will be listed together on the special issue website. Research articles, review articles as well as short communications are invited. For planned papers, a title and short abstract (about 100 words) can be sent to the Editorial Office for announcement on this website.

Submitted manuscripts should not have been published previously, nor be under consideration for publication elsewhere (except conference proceedings papers). All manuscripts are thoroughly refereed through a single-blind peer-review process. A guide for authors and other relevant information for submission of manuscripts is available on the Instructions for Authors page. Remote Sensing is an international peer-reviewed open access semimonthly journal published by MDPI.

Please visit the Instructions for Authors page before submitting a manuscript. The Article Processing Charge (APC) for publication in this open access journal is 2700 CHF (Swiss Francs). Submitted papers should be well formatted and use good English. Authors may use MDPI's English editing service prior to publication or during author revisions.

• CNNs
• RCNN
• deep reinforcement learning
• machine learning algorithms
• ensemble algorithms
• random forest
• support vector machines
• Gaussian Bayes
• decision tree classifier
• vegetation mapping, monitoring, trend analysis